[PATCH] ext2fs and large stores (> 1.5G)

[Ognyan Kulev]

I'm very pleased to announce the availability of an _alpha-quality_ 
patch that solves one of the most irritating inconveniences in the Hurd.

Getting the Patch
-----------------

http://debian.fmi.uni-sofia.bg/~ogi/hurd/ext3fs/

Using the Modified ext2fs
-------------------------

Apply the patch to the latest Hurd source and compile the ext2fs 
translator.  You can put it in /hurd/ext2fs-alfa, for example. 
Read-only mode should be stable.  Read-write mode can _corrupt data_. 
Always use ext2fs-alfa as active translator.

Bug Reports
-----------

When reporting bugs, please include the debug output of the translator 
(compressed with bzip2 -9).  Use the following command:

$ settrans -a test /hurd/ext2fs-alfa -D /dev/hdXsY 2> test.debug

Don't fill mailboxes with large bzip2 files.  Put them somewhere in 
Internet, or send them as private mail to me (ogi@fmi.uni-sofia.bg).

Patches are even better :-)

Design
------

The idea outlined in 
http://lists.debian.org/debian-hurd/2003/debian-hurd-200302/msg00016.html 
is used.  A very short description of the implementation follows.

To achieve this, ext2fs reads the superblock very early with store_read. 
 This allows the exact places of all the meta blocks at fixed locations 
to be calculated.  A mapping is created (with the help of 
disk_image_bptr  and disk_image_boffs) in the beginning of disk_image. 
(See disk_pager_init.)

The rest of disk_image is used as block cache for indirect blocks.  Code 
that uses indirect block must use disk_image_request and 
disk_image_release.  disk_image_request first checks if the block is 
already in the memory, and if it's not, then allocate it from the block 
cache.

When using disk_image_release for modified blocks, they are marked as 
"dirty".  The actual write of the block happens in the pokel or 
sync_global_ptr which call disk_image_clear.  disk_image_clear clears 
the dirty flag of the block thus allowing it to be reused.

To Do
-----

* block sizes different than 4096 are not supported.  Don't try it! 
Discussion about this issue is welcomed, as I didn't think about this yet.

Future
------

The beta release of the patch will solve the block size limit.

I would like to see ext2fs server without the 1G limit into the Hurd 
source tree.  If the core developers want changes to this patch, let's 
discuss them.

Misc.
-----

The implementation took 2 weeks.  It was easier than porting CMD649 
ATA-100 PCI driver to OSKit, and much more fun too :-)

Enjoy!
--
Ognyan Kulev <ogi@fmi.uni-sofia.bg>, "\"Programmer\""